ioremap and dma cleanups and fixes for superh (resend)

ioremap and dma cleanups and fixes for superh (resend)

[Christoph Hellwig]

Hi Yoshinori and Rich,

can you take a look and possibly pick up the series below that untangles
and sorts out minor issues with the sh ioremap and dma code?

[PATCH 01/10] sh: remove -Werror from Makefiles

[Christoph Hellwig]

The sh build is full of warnings when building with gcc 9.2.1.  While
fixing those would be great, at least avoid failing the build.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/sh/kernel/Makefile | 2 --
 arch/sh/lib/Makefile    | 2 --
 arch/sh/mm/Makefile     | 2 --
 3 files changed, 6 deletions(-)

diff --git a/arch/sh/kernel/Makefile b/arch/sh/kernel/Makefile
index b0f5574b6228b3..aa0fbc9202b14e 100644
--- a/arch/sh/kernel/Makefile
+++ b/arch/sh/kernel/Makefile
@@ -47,5 +47,3 @@ obj-$(CONFIG_DWARF_UNWINDER)	+= dwarf.o
 obj-$(CONFIG_PERF_EVENTS)	+= perf_event.o perf_callchain.o
 obj-$(CONFIG_DMA_NONCOHERENT)	+= dma-coherent.o
 obj-$(CONFIG_HAVE_HW_BREAKPOINT)		+= hw_breakpoint.o
-
-ccflags-y := -Werror
diff --git a/arch/sh/lib/Makefile b/arch/sh/lib/Makefile
index d0abbe5e38b0f3..eb473d373ca43a 100644
--- a/arch/sh/lib/Makefile
+++ b/arch/sh/lib/Makefile
@@ -30,5 +30,3 @@ memset-$(CONFIG_CPU_SH4)	:= memset-sh4.o
 lib-$(CONFIG_MMU)		+= copy_page.o __clear_user.o
 lib-$(CONFIG_MCOUNT)		+= mcount.o
 lib-y				+= $(memcpy-y) $(memset-y) $(udivsi3-y)
-
-ccflags-y := -Werror
diff --git a/arch/sh/mm/Makefile b/arch/sh/mm/Makefile
index 487da0ff03b3a7..f69ddc70b14658 100644
--- a/arch/sh/mm/Makefile
+++ b/arch/sh/mm/Makefile
@@ -43,5 +43,3 @@ obj-$(CONFIG_UNCACHED_MAPPING)	+= uncached.o
 obj-$(CONFIG_HAVE_SRAM_POOL)	+= sram.o
 
 GCOV_PROFILE_pmb.o := n
-
-ccflags-y := -Werror
-- 
2.26.2

[PATCH 02/10] sh: sort the selects for SUPERH alphabetically

[Christoph Hellwig]

Ensure there is an order for the selects.  Also remove a duplicate
one.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/sh/Kconfig | 95 ++++++++++++++++++++++++-------------------------
 1 file changed, 47 insertions(+), 48 deletions(-)

diff --git a/arch/sh/Kconfig b/arch/sh/Kconfig
index 9fc2b010e938c9..f8027eee08edae 100644
--- a/arch/sh/Kconfig
+++ b/arch/sh/Kconfig
@@ -1,75 +1,74 @@
 # SPDX-License-Identifier: GPL-2.0
 config SUPERH
 	def_bool y
+	select ARCH_32BIT_OFF_T
+	select ARCH_HAVE_CUSTOM_GPIO_H
+	select ARCH_HAVE_NMI_SAFE_CMPXCHG if (GUSA_RB || CPU_SH4A)
 	select ARCH_HAS_BINFMT_FLAT if !MMU
+	select ARCH_HAS_GIGANTIC_PAGE
+	select ARCH_HAS_GCOV_PROFILE_ALL
 	select ARCH_HAS_PTE_SPECIAL
 	select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
+	select ARCH_HIBERNATION_POSSIBLE if MMU
 	select ARCH_MIGHT_HAVE_PC_PARPORT
-	select HAVE_PATA_PLATFORM
+	select ARCH_WANT_IPC_PARSE_VERSION
 	select CLKDEV_LOOKUP
+	select CPU_NO_EFFICIENT_FFS
 	select DMA_DECLARE_COHERENT
-	select HAVE_IDE if HAS_IOPORT_MAP
-	select HAVE_OPROFILE
+	select GENERIC_ATOMIC64
+	select GENERIC_CLOCKEVENTS
+	select GENERIC_CMOS_UPDATE if SH_SH03 || SH_DREAMCAST
+	select GENERIC_IDLE_POLL_SETUP
+	select GENERIC_IRQ_SHOW
+	select GENERIC_PCI_IOMAP if PCI
+	select GENERIC_SCHED_CLOCK
+	select GENERIC_STRNCPY_FROM_USER
+	select GENERIC_STRNLEN_USER
+	select GENERIC_SMP_IDLE_THREAD
+	select GUP_GET_PTE_LOW_HIGH if X2TLB
+	select HAVE_ARCH_AUDITSYSCALL
+	select HAVE_ARCH_KGDB
 	select HAVE_ARCH_TRACEHOOK
-	select HAVE_PERF_EVENTS
 	select HAVE_DEBUG_BUGVERBOSE
-	select HAVE_FAST_GUP if MMU
-	select ARCH_HAVE_CUSTOM_GPIO_H
-	select ARCH_HAVE_NMI_SAFE_CMPXCHG if (GUSA_RB || CPU_SH4A)
-	select ARCH_HAS_GCOV_PROFILE_ALL
-	select PERF_USE_VMALLOC
 	select HAVE_DEBUG_KMEMLEAK
-	select HAVE_KERNEL_GZIP
-	select CPU_NO_EFFICIENT_FFS
+	select HAVE_DYNAMIC_FTRACE
+	select HAVE_FAST_GUP if MMU
+	select HAVE_FUNCTION_GRAPH_TRACER
+	select HAVE_FUNCTION_TRACER
+	select HAVE_FUTEX_CMPXCHG if FUTEX
+	select HAVE_FTRACE_MCOUNT_RECORD
+	select HAVE_HW_BREAKPOINT
+	select HAVE_IDE if HAS_IOPORT_MAP
+	select HAVE_IOREMAP_PROT if MMU && !X2TLB
 	select HAVE_KERNEL_BZIP2
+	select HAVE_KERNEL_GZIP
 	select HAVE_KERNEL_LZMA
-	select HAVE_KERNEL_XZ
 	select HAVE_KERNEL_LZO
+	select HAVE_KERNEL_XZ
+	select HAVE_KPROBES
+	select HAVE_KRETPROBES
+	select HAVE_MIXED_BREAKPOINTS_REGS
+	select HAVE_MOD_ARCH_SPECIFIC if DWARF_UNWINDER
+	select HAVE_NMI
+	select HAVE_OPROFILE
+	select HAVE_PATA_PLATFORM
+	select HAVE_PERF_EVENTS
+	select HAVE_REGS_AND_STACK_ACCESS_API
 	select HAVE_UID16
-	select ARCH_WANT_IPC_PARSE_VERSION
+	select HAVE_STACKPROTECTOR
 	select HAVE_SYSCALL_TRACEPOINTS
-	select HAVE_REGS_AND_STACK_ACCESS_API
-	select MAY_HAVE_SPARSE_IRQ
 	select IRQ_FORCED_THREADING
-	select RTC_LIB
-	select GENERIC_ATOMIC64
-	select GENERIC_IRQ_SHOW
-	select GENERIC_SMP_IDLE_THREAD
-	select GENERIC_IDLE_POLL_SETUP
-	select GENERIC_CLOCKEVENTS
-	select GENERIC_CMOS_UPDATE if SH_SH03 || SH_DREAMCAST
-	select GENERIC_PCI_IOMAP if PCI
-	select GENERIC_SCHED_CLOCK
-	select GENERIC_STRNCPY_FROM_USER
-	select GENERIC_STRNLEN_USER
-	select HAVE_MOD_ARCH_SPECIFIC if DWARF_UNWINDER
+	select MAY_HAVE_SPARSE_IRQ
 	select MODULES_USE_ELF_RELA
+	select NEED_SG_DMA_LENGTH
 	select NO_GENERIC_PCI_IOPORT_MAP if PCI
-	select OLD_SIGSUSPEND
 	select OLD_SIGACTION
+	select OLD_SIGSUSPEND
 	select PCI_DOMAINS if PCI
-	select HAVE_ARCH_AUDITSYSCALL
-	select HAVE_FUTEX_CMPXCHG if FUTEX
-	select HAVE_NMI
-	select NEED_SG_DMA_LENGTH
-	select ARCH_HAS_GIGANTIC_PAGE
-	select ARCH_32BIT_OFF_T
-	select GUP_GET_PTE_LOW_HIGH if X2TLB
-	select HAVE_KPROBES
-	select HAVE_KRETPROBES
-	select HAVE_IOREMAP_PROT if MMU && !X2TLB
-	select HAVE_FUNCTION_TRACER
-	select HAVE_FTRACE_MCOUNT_RECORD
-	select HAVE_DYNAMIC_FTRACE
-	select ARCH_WANT_IPC_PARSE_VERSION
-	select HAVE_FUNCTION_GRAPH_TRACER
-	select HAVE_ARCH_KGDB
-	select HAVE_HW_BREAKPOINT
-	select HAVE_MIXED_BREAKPOINTS_REGS
 	select PERF_EVENTS
-	select ARCH_HIBERNATION_POSSIBLE if MMU
+	select PERF_USE_VMALLOC
+	select RTC_LIB
 	select SPARSE_IRQ
-	select HAVE_STACKPROTECTOR
 	help
 	  The SuperH is a RISC processor targeted for use in embedded systems
 	  and consumer electronics; it was also used in the Sega Dreamcast
-- 
2.26.2

[PATCH 03/10] sh: remove __KERNEL__ ifdefs from non-UAPI headers

[Christoph Hellwig]

There is no point in having __KERNEL__ ifdefs in headers not exported to
userspace.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/sh/include/asm/adc.h          | 2 --
 arch/sh/include/asm/addrspace.h    | 3 ---
 arch/sh/include/asm/bitops.h       | 4 ----
 arch/sh/include/asm/cache.h        | 2 --
 arch/sh/include/asm/cacheflush.h   | 3 ---
 arch/sh/include/asm/dma.h          | 2 --
 arch/sh/include/asm/elf.h          | 2 --
 arch/sh/include/asm/freq.h         | 2 --
 arch/sh/include/asm/futex.h        | 3 ---
 arch/sh/include/asm/io.h           | 3 ---
 arch/sh/include/asm/mmu_context.h  | 2 --
 arch/sh/include/asm/mmzone.h       | 3 ---
 arch/sh/include/asm/pci.h          | 4 ----
 arch/sh/include/asm/processor_32.h | 2 --
 arch/sh/include/asm/sparsemem.h    | 3 ---
 arch/sh/include/asm/string_32.h    | 4 ----
 arch/sh/include/asm/syscalls_32.h  | 3 ---
 arch/sh/include/asm/thread_info.h  | 5 -----
 arch/sh/include/asm/watchdog.h     | 2 --
 19 files changed, 54 deletions(-)

diff --git a/arch/sh/include/asm/adc.h b/arch/sh/include/asm/adc.h
index 99ec668495594c..feccfe639e3896 100644
--- a/arch/sh/include/asm/adc.h
+++ b/arch/sh/include/asm/adc.h
@@ -1,7 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 #ifndef __ASM_ADC_H
 #define __ASM_ADC_H
-#ifdef __KERNEL__
 /*
  * Copyright (C) 2004  Andriy Skulysh
  */
@@ -10,5 +9,4 @@
 
 int adc_single(unsigned int channel);
 
-#endif /* __KERNEL__ */
 #endif /* __ASM_ADC_H */
diff --git a/arch/sh/include/asm/addrspace.h b/arch/sh/include/asm/addrspace.h
index 34bfbcddcce027..468fba333e896b 100644
--- a/arch/sh/include/asm/addrspace.h
+++ b/arch/sh/include/asm/addrspace.h
@@ -7,8 +7,6 @@
 #ifndef __ASM_SH_ADDRSPACE_H
 #define __ASM_SH_ADDRSPACE_H
 
-#ifdef __KERNEL__
-
 #include <cpu/addrspace.h>
 
 /* If this CPU supports segmentation, hook up the helpers */
@@ -62,5 +60,4 @@
 #define P3_ADDR_MAX		P4SEG
 #endif
 
-#endif /* __KERNEL__ */
 #endif /* __ASM_SH_ADDRSPACE_H */
diff --git a/arch/sh/include/asm/bitops.h b/arch/sh/include/asm/bitops.h
index 445dd14c448a67..450b5854d7c67b 100644
--- a/arch/sh/include/asm/bitops.h
+++ b/arch/sh/include/asm/bitops.h
@@ -2,8 +2,6 @@
 #ifndef __ASM_SH_BITOPS_H
 #define __ASM_SH_BITOPS_H
 
-#ifdef __KERNEL__
-
 #ifndef _LINUX_BITOPS_H
 #error only <linux/bitops.h> can be included directly
 #endif
@@ -71,6 +69,4 @@ static inline unsigned long __ffs(unsigned long word)
 #include <asm-generic/bitops/__fls.h>
 #include <asm-generic/bitops/fls64.h>
 
-#endif /* __KERNEL__ */
-
 #endif /* __ASM_SH_BITOPS_H */
diff --git a/arch/sh/include/asm/cache.h b/arch/sh/include/asm/cache.h
index 2408ac4873aa00..a293343456affe 100644
--- a/arch/sh/include/asm/cache.h
+++ b/arch/sh/include/asm/cache.h
@@ -8,7 +8,6 @@
  */
 #ifndef __ASM_SH_CACHE_H
 #define __ASM_SH_CACHE_H
-#ifdef __KERNEL__
 
 #include <linux/init.h>
 #include <cpu/cache.h>
@@ -44,5 +43,4 @@ struct cache_info {
 	unsigned long flags;
 };
 #endif /* __ASSEMBLY__ */
-#endif /* __KERNEL__ */
 #endif /* __ASM_SH_CACHE_H */
diff --git a/arch/sh/include/asm/cacheflush.h b/arch/sh/include/asm/cacheflush.h
index fe7400079b97b0..4486a865ff62f6 100644
--- a/arch/sh/include/asm/cacheflush.h
+++ b/arch/sh/include/asm/cacheflush.h
@@ -2,8 +2,6 @@
 #ifndef __ASM_SH_CACHEFLUSH_H
 #define __ASM_SH_CACHEFLUSH_H
 
-#ifdef __KERNEL__
-
 #include <linux/mm.h>
 
 /*
@@ -109,5 +107,4 @@ static inline void *sh_cacheop_vaddr(void *vaddr)
 	return vaddr;
 }
 
-#endif /* __KERNEL__ */
 #endif /* __ASM_SH_CACHEFLUSH_H */
diff --git a/arch/sh/include/asm/dma.h b/arch/sh/include/asm/dma.h
index 4d5a21a891c078..17d23ae98c77a4 100644
--- a/arch/sh/include/asm/dma.h
+++ b/arch/sh/include/asm/dma.h
@@ -6,7 +6,6 @@
  */
 #ifndef __ASM_SH_DMA_H
 #define __ASM_SH_DMA_H
-#ifdef __KERNEL__
 
 #include <linux/spinlock.h>
 #include <linux/wait.h>
@@ -144,5 +143,4 @@ extern int isa_dma_bridge_buggy;
 #define isa_dma_bridge_buggy	(0)
 #endif
 
-#endif /* __KERNEL__ */
 #endif /* __ASM_SH_DMA_H */
diff --git a/arch/sh/include/asm/elf.h b/arch/sh/include/asm/elf.h
index 7661fb5d548a08..2862d6d1cb64f1 100644
--- a/arch/sh/include/asm/elf.h
+++ b/arch/sh/include/asm/elf.h
@@ -90,7 +90,6 @@ typedef struct user_fpu_struct elf_fpregset_t;
 #endif
 #define ELF_ARCH	EM_SH
 
-#ifdef __KERNEL__
 /*
  * This is used to ensure we don't load something for the wrong architecture.
  */
@@ -209,5 +208,4 @@ do {								\
 	NEW_AUX_ENT(AT_L2_CACHESHAPE, l2_cache_shape);		\
 } while (0)
 
-#endif /* __KERNEL__ */
 #endif /* __ASM_SH_ELF_H */
diff --git a/arch/sh/include/asm/freq.h b/arch/sh/include/asm/freq.h
index 18133bf837386b..87c23621b7ae81 100644
--- a/arch/sh/include/asm/freq.h
+++ b/arch/sh/include/asm/freq.h
@@ -6,9 +6,7 @@
  */
 #ifndef __ASM_SH_FREQ_H
 #define __ASM_SH_FREQ_H
-#ifdef __KERNEL__
 
 #include <cpu/freq.h>
 
-#endif /* __KERNEL__ */
 #endif /* __ASM_SH_FREQ_H */
diff --git a/arch/sh/include/asm/futex.h b/arch/sh/include/asm/futex.h
index b39cda09fb95cc..b70f3fce6ed77a 100644
--- a/arch/sh/include/asm/futex.h
+++ b/arch/sh/include/asm/futex.h
@@ -2,8 +2,6 @@
 #ifndef __ASM_SH_FUTEX_H
 #define __ASM_SH_FUTEX_H
 
-#ifdef __KERNEL__
-
 #include <linux/futex.h>
 #include <linux/uaccess.h>
 #include <asm/errno.h>
@@ -71,5 +69,4 @@ static inline int arch_futex_atomic_op_inuser(int op, u32 oparg, int *oval,
 	return ret;
 }
 
-#endif /* __KERNEL__ */
 #endif /* __ASM_SH_FUTEX_H */
diff --git a/arch/sh/include/asm/io.h b/arch/sh/include/asm/io.h
index 26f0f9b4658b20..1fd06ef6a19454 100644
--- a/arch/sh/include/asm/io.h
+++ b/arch/sh/include/asm/io.h
@@ -20,7 +20,6 @@
 #include <linux/pgtable.h>
 #include <asm-generic/iomap.h>
 
-#ifdef __KERNEL__
 #define __IO_PREFIX     generic
 #include <asm/io_generic.h>
 #include <asm/io_trapped.h>
@@ -380,6 +379,4 @@ static inline int iounmap_fixed(void __iomem *addr) { return -EINVAL; }
 int valid_phys_addr_range(phys_addr_t addr, size_t size);
 int valid_mmap_phys_addr_range(unsigned long pfn, size_t size);
 
-#endif /* __KERNEL__ */
-
 #endif /* __ASM_SH_IO_H */
diff --git a/arch/sh/include/asm/mmu_context.h b/arch/sh/include/asm/mmu_context.h
index 48e67d544d53cf..f664e51e8a15e5 100644
--- a/arch/sh/include/asm/mmu_context.h
+++ b/arch/sh/include/asm/mmu_context.h
@@ -8,7 +8,6 @@
 #ifndef __ASM_SH_MMU_CONTEXT_H
 #define __ASM_SH_MMU_CONTEXT_H
 
-#ifdef __KERNEL__
 #include <cpu/mmu_context.h>
 #include <asm/tlbflush.h>
 #include <linux/uaccess.h>
@@ -177,5 +176,4 @@ static inline void disable_mmu(void)
 #define disable_mmu()	do { } while (0)
 #endif
 
-#endif /* __KERNEL__ */
 #endif /* __ASM_SH_MMU_CONTEXT_H */
diff --git a/arch/sh/include/asm/mmzone.h b/arch/sh/include/asm/mmzone.h
index cbaee1d1b67320..6552a088dc97ae 100644
--- a/arch/sh/include/asm/mmzone.h
+++ b/arch/sh/include/asm/mmzone.h
@@ -2,8 +2,6 @@
 #ifndef __ASM_SH_MMZONE_H
 #define __ASM_SH_MMZONE_H
 
-#ifdef __KERNEL__
-
 #ifdef CONFIG_NEED_MULTIPLE_NODES
 #include <linux/numa.h>
 
@@ -44,5 +42,4 @@ void __init __add_active_range(unsigned int nid, unsigned long start_pfn,
 /* arch/sh/mm/init.c */
 void __init allocate_pgdat(unsigned int nid);
 
-#endif /* __KERNEL__ */
 #endif /* __ASM_SH_MMZONE_H */
diff --git a/arch/sh/include/asm/pci.h b/arch/sh/include/asm/pci.h
index 10a36b1cf2eab1..ad22e88c6657dc 100644
--- a/arch/sh/include/asm/pci.h
+++ b/arch/sh/include/asm/pci.h
@@ -2,8 +2,6 @@
 #ifndef __ASM_SH_PCI_H
 #define __ASM_SH_PCI_H
 
-#ifdef __KERNEL__
-
 /* Can be used to override the logic in pci_scan_bus for skipping
    already-configured bus numbers - to be used for buggy BIOSes
    or architectures with incomplete PCI setup by the loader */
@@ -96,6 +94,4 @@ static inline int pci_get_legacy_ide_irq(struct pci_dev *dev, int channel)
 	return channel ? 15 : 14;
 }
 
-#endif /* __KERNEL__ */
 #endif /* __ASM_SH_PCI_H */
-
diff --git a/arch/sh/include/asm/processor_32.h b/arch/sh/include/asm/processor_32.h
index d44409413418a5..aa92cc933889df 100644
--- a/arch/sh/include/asm/processor_32.h
+++ b/arch/sh/include/asm/processor_32.h
@@ -8,7 +8,6 @@
 
 #ifndef __ASM_SH_PROCESSOR_32_H
 #define __ASM_SH_PROCESSOR_32_H
-#ifdef __KERNEL__
 
 #include <linux/compiler.h>
 #include <linux/linkage.h>
@@ -203,5 +202,4 @@ static inline void prefetchw(const void *x)
 }
 #endif
 
-#endif /* __KERNEL__ */
 #endif /* __ASM_SH_PROCESSOR_32_H */
diff --git a/arch/sh/include/asm/sparsemem.h b/arch/sh/include/asm/sparsemem.h
index 4eb899751e4590..ed287c4980bcab 100644
--- a/arch/sh/include/asm/sparsemem.h
+++ b/arch/sh/include/asm/sparsemem.h
@@ -2,7 +2,6 @@
 #ifndef __ASM_SH_SPARSEMEM_H
 #define __ASM_SH_SPARSEMEM_H
 
-#ifdef __KERNEL__
 /*
  * SECTION_SIZE_BITS		2^N: how big each section will be
  * MAX_PHYSADDR_BITS		2^N: how much physical address space we have
@@ -12,6 +11,4 @@
 #define MAX_PHYSADDR_BITS	32
 #define MAX_PHYSMEM_BITS	32
 
-#endif
-
 #endif /* __ASM_SH_SPARSEMEM_H */
diff --git a/arch/sh/include/asm/string_32.h b/arch/sh/include/asm/string_32.h
index 3558b1d7123e9c..778589e4a99dbb 100644
--- a/arch/sh/include/asm/string_32.h
+++ b/arch/sh/include/asm/string_32.h
@@ -2,8 +2,6 @@
 #ifndef __ASM_SH_STRING_H
 #define __ASM_SH_STRING_H
 
-#ifdef __KERNEL__
-
 /*
  * Copyright (C) 1999 Niibe Yutaka
  * But consider these trivial functions to be public domain.
@@ -127,6 +125,4 @@ extern void *memchr(const void *__s, int __c, size_t __n);
 #define __HAVE_ARCH_STRLEN
 extern size_t strlen(const char *);
 
-#endif /* __KERNEL__ */
-
 #endif /* __ASM_SH_STRING_H */
diff --git a/arch/sh/include/asm/syscalls_32.h b/arch/sh/include/asm/syscalls_32.h
index 9f9faf63b48c2e..5c555b864fe092 100644
--- a/arch/sh/include/asm/syscalls_32.h
+++ b/arch/sh/include/asm/syscalls_32.h
@@ -2,8 +2,6 @@
 #ifndef __ASM_SH_SYSCALLS_32_H
 #define __ASM_SH_SYSCALLS_32_H
 
-#ifdef __KERNEL__
-
 #include <linux/compiler.h>
 #include <linux/linkage.h>
 #include <linux/types.h>
@@ -26,5 +24,4 @@ asmlinkage void do_syscall_trace_leave(struct pt_regs *regs);
 asmlinkage void do_notify_resume(struct pt_regs *regs, unsigned int save_r0,
 				 unsigned long thread_info_flags);
 
-#endif /* __KERNEL__ */
 #endif /* __ASM_SH_SYSCALLS_32_H */
diff --git a/arch/sh/include/asm/thread_info.h b/arch/sh/include/asm/thread_info.h
index 6404be69d5fa6b..243ea5150aa00b 100644
--- a/arch/sh/include/asm/thread_info.h
+++ b/arch/sh/include/asm/thread_info.h
@@ -10,8 +10,6 @@
  *  Copyright (C) 2002  David Howells (dhowells@redhat.com)
  *  - Incorporating suggestions made by Linus Torvalds and Dave Miller
  */
-#ifdef __KERNEL__
-
 #include <asm/page.h>
 
 /*
@@ -170,7 +168,4 @@ static inline unsigned int get_thread_fault_code(void)
 }
 
 #endif	/* !__ASSEMBLY__ */
-
-#endif /* __KERNEL__ */
-
 #endif /* __ASM_SH_THREAD_INFO_H */
diff --git a/arch/sh/include/asm/watchdog.h b/arch/sh/include/asm/watchdog.h
index cecd0fc507f938..b9ca4c99f046a6 100644
--- a/arch/sh/include/asm/watchdog.h
+++ b/arch/sh/include/asm/watchdog.h
@@ -8,7 +8,6 @@
  */
 #ifndef __ASM_SH_WATCHDOG_H
 #define __ASM_SH_WATCHDOG_H
-#ifdef __KERNEL__
 
 #include <linux/types.h>
 #include <linux/io.h>
@@ -157,5 +156,4 @@ static inline void sh_wdt_write_csr(__u8 val)
 	__raw_writew((WTCSR_HIGH << 8) | (__u16)val, WTCSR);
 }
 #endif /* CONFIG_CPU_SUBTYPE_SH7785 || CONFIG_CPU_SUBTYPE_SH7780 */
-#endif /* __KERNEL__ */
 #endif /* __ASM_SH_WATCHDOG_H */
-- 
2.26.2

[PATCH 04/10] sh: move ioremap_fixed details out of <asm/io.h>

[Christoph Hellwig]

ioremap_fixed is an internal implementation detail and should not be
exposed to drivers.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/sh/include/asm/io.h   | 16 ----------------
 arch/sh/mm/init.c          |  1 +
 arch/sh/mm/ioremap.c       |  1 +
 arch/sh/mm/ioremap.h       | 23 +++++++++++++++++++++++
 arch/sh/mm/ioremap_fixed.c |  1 +
 5 files changed, 26 insertions(+), 16 deletions(-)
 create mode 100644 arch/sh/mm/ioremap.h

diff --git a/arch/sh/include/asm/io.h b/arch/sh/include/asm/io.h
index 1fd06ef6a19454..357a7e0c86d682 100644
--- a/arch/sh/include/asm/io.h
+++ b/arch/sh/include/asm/io.h
@@ -346,22 +346,6 @@ ioremap_prot(phys_addr_t offset, unsigned long size, unsigned long flags)
 }
 #endif
 
-#ifdef CONFIG_IOREMAP_FIXED
-extern void __iomem *ioremap_fixed(phys_addr_t, unsigned long, pgprot_t);
-extern int iounmap_fixed(void __iomem *);
-extern void ioremap_fixed_init(void);
-#else
-static inline void __iomem *
-ioremap_fixed(phys_addr_t phys_addr, unsigned long size, pgprot_t prot)
-{
-	BUG();
-	return NULL;
-}
-
-static inline void ioremap_fixed_init(void) { }
-static inline int iounmap_fixed(void __iomem *addr) { return -EINVAL; }
-#endif
-
 #define ioremap_uc	ioremap
 
 /*
diff --git a/arch/sh/mm/init.c b/arch/sh/mm/init.c
index a70ba0fdd0b382..da7ea48f9439e2 100644
--- a/arch/sh/mm/init.c
+++ b/arch/sh/mm/init.c
@@ -28,6 +28,7 @@
 #include <asm/setup.h>
 #include <asm/cache.h>
 #include <linux/sizes.h>
+#include "ioremap.h"
 
 pgd_t swapper_pg_dir[PTRS_PER_PGD];
 
diff --git a/arch/sh/mm/ioremap.c b/arch/sh/mm/ioremap.c
index f6d02246d665bd..d9ec85b6bb2130 100644
--- a/arch/sh/mm/ioremap.c
+++ b/arch/sh/mm/ioremap.c
@@ -24,6 +24,7 @@
 #include <asm/cacheflush.h>
 #include <asm/tlbflush.h>
 #include <asm/mmu.h>
+#include "ioremap.h"
 
 /*
  * Remap an arbitrary physical address space into the kernel virtual
diff --git a/arch/sh/mm/ioremap.h b/arch/sh/mm/ioremap.h
new file mode 100644
index 00000000000000..f2544e721a3526
--- /dev/null
+++ b/arch/sh/mm/ioremap.h
@@ -0,0 +1,23 @@
+#ifndef _SH_MM_IORMEMAP_H
+#define _SH_MM_IORMEMAP_H 1
+
+#ifdef CONFIG_IOREMAP_FIXED
+void __iomem *ioremap_fixed(phys_addr_t, unsigned long, pgprot_t);
+int iounmap_fixed(void __iomem *);
+void ioremap_fixed_init(void);
+#else
+static inline void __iomem *
+ioremap_fixed(phys_addr_t phys_addr, unsigned long size, pgprot_t prot)
+{
+	BUG();
+	return NULL;
+}
+static inline void ioremap_fixed_init(void)
+{
+}
+static inline int iounmap_fixed(void __iomem *addr)
+{
+	return -EINVAL;
+}
+#endif /* CONFIG_IOREMAP_FIXED */
+#endif /* _SH_MM_IORMEMAP_H */
diff --git a/arch/sh/mm/ioremap_fixed.c b/arch/sh/mm/ioremap_fixed.c
index 07e744d75fa01b..1914b79d1c530b 100644
--- a/arch/sh/mm/ioremap_fixed.c
+++ b/arch/sh/mm/ioremap_fixed.c
@@ -24,6 +24,7 @@
 #include <asm/tlbflush.h>
 #include <asm/mmu.h>
 #include <asm/mmu_context.h>
+#include "ioremap.h"
 
 struct ioremap_map {
 	void __iomem *addr;
-- 
2.26.2

[PATCH 05/10] sh: move the ioremap implementation out of line

[Christoph Hellwig]

Move the internal implementation details of ioremap out of line, no need
to expose any of this to drivers for a slow path API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/sh/include/asm/io.h | 101 ++++++---------------------------------
 arch/sh/mm/ioremap.c     |  53 ++++++++++++++++++++
 2 files changed, 68 insertions(+), 86 deletions(-)

diff --git a/arch/sh/include/asm/io.h b/arch/sh/include/asm/io.h
index 357a7e0c86d682..da08a61a2f7dae 100644
--- a/arch/sh/include/asm/io.h
+++ b/arch/sh/include/asm/io.h
@@ -242,109 +242,38 @@ unsigned long long poke_real_address_q(unsigned long long addr,
 #define phys_to_virt(address)	(__va(address))
 #endif
 
-/*
- * On 32-bit SH, we traditionally have the whole physical address space
- * mapped at all times (as MIPS does), so "ioremap()" and "iounmap()" do
- * not need to do anything but place the address in the proper segment.
- * This is true for P1 and P2 addresses, as well as some P3 ones.
- * However, most of the P3 addresses and newer cores using extended
- * addressing need to map through page tables, so the ioremap()
- * implementation becomes a bit more complicated.
- *
- * See arch/sh/mm/ioremap.c for additional notes on this.
- *
- * We cheat a bit and always return uncachable areas until we've fixed
- * the drivers to handle caching properly.
- *
- * On the SH-5 the concept of segmentation in the 1:1 PXSEG sense simply
- * doesn't exist, so everything must go through page tables.
- */
 #ifdef CONFIG_MMU
+void iounmap(void __iomem *addr);
 void __iomem *__ioremap_caller(phys_addr_t offset, unsigned long size,
 			       pgprot_t prot, void *caller);
-void iounmap(void __iomem *addr);
-
-static inline void __iomem *
-__ioremap(phys_addr_t offset, unsigned long size, pgprot_t prot)
-{
-	return __ioremap_caller(offset, size, prot, __builtin_return_address(0));
-}
-
-static inline void __iomem *
-__ioremap_29bit(phys_addr_t offset, unsigned long size, pgprot_t prot)
-{
-#ifdef CONFIG_29BIT
-	phys_addr_t last_addr = offset + size - 1;
-
-	/*
-	 * For P1 and P2 space this is trivial, as everything is already
-	 * mapped. Uncached access for P1 addresses are done through P2.
-	 * In the P3 case or for addresses outside of the 29-bit space,
-	 * mapping must be done by the PMB or by using page tables.
-	 */
-	if (likely(PXSEG(offset) < P3SEG && PXSEG(last_addr) < P3SEG)) {
-		u64 flags = pgprot_val(prot);
-
-		/*
-		 * Anything using the legacy PTEA space attributes needs
-		 * to be kicked down to page table mappings.
-		 */
-		if (unlikely(flags & _PAGE_PCC_MASK))
-			return NULL;
-		if (unlikely(flags & _PAGE_CACHABLE))
-			return (void __iomem *)P1SEGADDR(offset);
-
-		return (void __iomem *)P2SEGADDR(offset);
-	}
-
-	/* P4 above the store queues are always mapped. */
-	if (unlikely(offset >= P3_ADDR_MAX))
-		return (void __iomem *)P4SEGADDR(offset);
-#endif
-
-	return NULL;
-}
-
-static inline void __iomem *
-__ioremap_mode(phys_addr_t offset, unsigned long size, pgprot_t prot)
-{
-	void __iomem *ret;
-
-	ret = __ioremap_trapped(offset, size);
-	if (ret)
-		return ret;
-
-	ret = __ioremap_29bit(offset, size, prot);
-	if (ret)
-		return ret;
-
-	return __ioremap(offset, size, prot);
-}
-#else
-#define __ioremap(offset, size, prot)		((void __iomem *)(offset))
-#define __ioremap_mode(offset, size, prot)	((void __iomem *)(offset))
-static inline void iounmap(void __iomem *addr) {}
-#endif /* CONFIG_MMU */
 
 static inline void __iomem *ioremap(phys_addr_t offset, unsigned long size)
 {
-	return __ioremap_mode(offset, size, PAGE_KERNEL_NOCACHE);
+	return __ioremap_caller(offset, size, PAGE_KERNEL_NOCACHE,
+			__builtin_return_address(0));
 }
 
 static inline void __iomem *
 ioremap_cache(phys_addr_t offset, unsigned long size)
 {
-	return __ioremap_mode(offset, size, PAGE_KERNEL);
+	return __ioremap_caller(offset, size, PAGE_KERNEL,
+			__builtin_return_address(0));
 }
 #define ioremap_cache ioremap_cache
 
 #ifdef CONFIG_HAVE_IOREMAP_PROT
-static inline void __iomem *
-ioremap_prot(phys_addr_t offset, unsigned long size, unsigned long flags)
+static inline void __iomem *ioremap_prot(phys_addr_t offset, unsigned long size,
+		unsigned long flags)
 {
-	return __ioremap_mode(offset, size, __pgprot(flags));
+	return __ioremap_caller(offset, size, __pgprot(flags),
+			__builtin_return_address(0));
 }
-#endif
+#endif /* CONFIG_HAVE_IOREMAP_PROT */
+
+#else /* CONFIG_MMU */
+#define iounmap(addr)		do { } while (0)
+#define ioremap(offset, size)	((void __iomem *)(unsigned long)(offset))
+#endif /* CONFIG_MMU */
 
 #define ioremap_uc	ioremap
 
diff --git a/arch/sh/mm/ioremap.c b/arch/sh/mm/ioremap.c
index d9ec85b6bb2130..69e55939e48a6f 100644
--- a/arch/sh/mm/ioremap.c
+++ b/arch/sh/mm/ioremap.c
@@ -26,6 +26,51 @@
 #include <asm/mmu.h>
 #include "ioremap.h"
 
+/*
+ * On 32-bit SH, we traditionally have the whole physical address space mapped
+ * at all times (as MIPS does), so "ioremap()" and "iounmap()" do not need to do
+ * anything but place the address in the proper segment.  This is true for P1
+ * and P2 addresses, as well as some P3 ones.  However, most of the P3 addresses
+ * and newer cores using extended addressing need to map through page tables, so
+ * the ioremap() implementation becomes a bit more complicated.
+ */
+#ifdef CONFIG_29BIT
+static void __iomem *
+__ioremap_29bit(phys_addr_t offset, unsigned long size, pgprot_t prot)
+{
+	phys_addr_t last_addr = offset + size - 1;
+
+	/*
+	 * For P1 and P2 space this is trivial, as everything is already
+	 * mapped. Uncached access for P1 addresses are done through P2.
+	 * In the P3 case or for addresses outside of the 29-bit space,
+	 * mapping must be done by the PMB or by using page tables.
+	 */
+	if (likely(PXSEG(offset) < P3SEG && PXSEG(last_addr) < P3SEG)) {
+		u64 flags = pgprot_val(prot);
+
+		/*
+		 * Anything using the legacy PTEA space attributes needs
+		 * to be kicked down to page table mappings.
+		 */
+		if (unlikely(flags & _PAGE_PCC_MASK))
+			return NULL;
+		if (unlikely(flags & _PAGE_CACHABLE))
+			return (void __iomem *)P1SEGADDR(offset);
+
+		return (void __iomem *)P2SEGADDR(offset);
+	}
+
+	/* P4 above the store queues are always mapped. */
+	if (unlikely(offset >= P3_ADDR_MAX))
+		return (void __iomem *)P4SEGADDR(offset);
+
+	return NULL;
+}
+#else
+#define __ioremap_29bit(offset, size, prot)		NULL
+#endif /* CONFIG_29BIT */
+
 /*
  * Remap an arbitrary physical address space into the kernel virtual
  * address space. Needed when the kernel wants to access high addresses
@@ -43,6 +88,14 @@ __ioremap_caller(phys_addr_t phys_addr, unsigned long size,
 	unsigned long offset, last_addr, addr, orig_addr;
 	void __iomem *mapped;
 
+	mapped = __ioremap_trapped(phys_addr, size);
+	if (mapped)
+		return mapped;
+
+	mapped = __ioremap_29bit(phys_addr, size, pgprot);
+	if (mapped)
+		return mapped;
+
 	/* Don't allow wraparound or zero size */
 	last_addr = phys_addr + size - 1;
 	if (!size || last_addr < phys_addr)
-- 
2.26.2

[PATCH 06/10] sh: don't include <asm/io_trapped.h> in <asm/io.h>

[Christoph Hellwig]

No need to expose the details of trapped I/O to drivers.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/sh/include/asm/io.h | 1 -
 arch/sh/kernel/ioport.c  | 1 +
 arch/sh/mm/ioremap.c     | 1 +
 3 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/sh/include/asm/io.h b/arch/sh/include/asm/io.h
index da08a61a2f7dae..873f1399fa5962 100644
--- a/arch/sh/include/asm/io.h
+++ b/arch/sh/include/asm/io.h
@@ -22,7 +22,6 @@
 
 #define __IO_PREFIX     generic
 #include <asm/io_generic.h>
-#include <asm/io_trapped.h>
 #include <asm-generic/pci_iomap.h>
 #include <mach/mangle-port.h>
 
diff --git a/arch/sh/kernel/ioport.c b/arch/sh/kernel/ioport.c
index 34f8cdbbcf0bee..f39446a658bdb8 100644
--- a/arch/sh/kernel/ioport.c
+++ b/arch/sh/kernel/ioport.c
@@ -7,6 +7,7 @@
  */
 #include <linux/module.h>
 #include <linux/io.h>
+#include <asm/io_trapped.h>
 
 unsigned long sh_io_port_base __read_mostly = -1;
 EXPORT_SYMBOL(sh_io_port_base);
diff --git a/arch/sh/mm/ioremap.c b/arch/sh/mm/ioremap.c
index 69e55939e48a6f..21342581144dee 100644
--- a/arch/sh/mm/ioremap.c
+++ b/arch/sh/mm/ioremap.c
@@ -18,6 +18,7 @@
 #include <linux/mm.h>
 #include <linux/pci.h>
 #include <linux/io.h>
+#include <asm/io_trapped.h>
 #include <asm/page.h>
 #include <asm/pgalloc.h>
 #include <asm/addrspace.h>
-- 
2.26.2

[PATCH 07/10] sh: unexport register_trapped_io and match_trapped_io_handler

[Christoph Hellwig]

Both functions are only used by compiled in core code.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/sh/kernel/io_trapped.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/arch/sh/kernel/io_trapped.c b/arch/sh/kernel/io_trapped.c
index 037aab2708b7ac..004ad0130b1038 100644
--- a/arch/sh/kernel/io_trapped.c
+++ b/arch/sh/kernel/io_trapped.c
@@ -102,7 +102,6 @@ int register_trapped_io(struct trapped_io *tiop)
 	pr_warn("unable to install trapped io filter\n");
 	return -1;
 }
-EXPORT_SYMBOL_GPL(register_trapped_io);
 
 void __iomem *match_trapped_io_handler(struct list_head *list,
 				       unsigned long offset,
@@ -131,7 +130,6 @@ void __iomem *match_trapped_io_handler(struct list_head *list,
 	spin_unlock_irqrestore(&trapped_lock, flags);
 	return NULL;
 }
-EXPORT_SYMBOL_GPL(match_trapped_io_handler);
 
 static struct trapped_io *lookup_tiop(unsigned long address)
 {
-- 
2.26.2

[PATCH 08/10] dma-mapping: consolidate the NO_DMA definition in kernel/dma/Kconfig

[Christoph Hellwig]

Have a single definition that architetures can select.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/m68k/Kconfig         | 4 +---
 arch/m68k/Kconfig.machine | 1 +
 arch/um/Kconfig           | 4 +---
 kernel/dma/Kconfig        | 3 +++
 4 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/arch/m68k/Kconfig b/arch/m68k/Kconfig
index 6ad6cdac74b3dc..8e488369a7e55a 100644
--- a/arch/m68k/Kconfig
+++ b/arch/m68k/Kconfig
@@ -17,6 +17,7 @@ config M68K
 	select HAVE_COPY_THREAD_TLS
 	select GENERIC_IRQ_SHOW
 	select GENERIC_ATOMIC64
+	select NO_DMA if !MMU && !COLDFIRE
 	select HAVE_UID16
 	select VIRT_TO_BUS
 	select ARCH_HAVE_NMI_SAFE_CMPXCHG if RMW_INSNS
@@ -60,9 +61,6 @@ config TIME_LOW_RES
 config NO_IOPORT_MAP
 	def_bool y
 
-config NO_DMA
-	def_bool (MMU && SUN3) || (!MMU && !COLDFIRE)
-
 config ZONE_DMA
 	bool
 	default y
diff --git a/arch/m68k/Kconfig.machine b/arch/m68k/Kconfig.machine
index a82651d58af483..17e8c3a292d770 100644
--- a/arch/m68k/Kconfig.machine
+++ b/arch/m68k/Kconfig.machine
@@ -126,6 +126,7 @@ config SUN3
 	depends on MMU
 	depends on !MMU_MOTOROLA
 	select MMU_SUN3 if MMU
+	select NO_DMA
 	select M68020
 	help
 	  This option enables support for the Sun 3 series of workstations
diff --git a/arch/um/Kconfig b/arch/um/Kconfig
index 9318dc6d1a0cee..32c1d1945033ec 100644
--- a/arch/um/Kconfig
+++ b/arch/um/Kconfig
@@ -15,6 +15,7 @@ config UML
 	select HAVE_DEBUG_KMEMLEAK
 	select HAVE_DEBUG_BUGVERBOSE
 	select HAVE_COPY_THREAD_TLS
+	select NO_DMA
 	select GENERIC_IRQ_SHOW
 	select GENERIC_CPU_DEVICES
 	select GENERIC_CLOCKEVENTS
@@ -168,9 +169,6 @@ config MMAPPER
 	  This driver allows a host file to be used as emulated IO memory inside
 	  UML.
 
-config NO_DMA
-	def_bool y
-
 config PGTABLE_LEVELS
 	int
 	default 3 if 3_LEVEL_PGTABLES
diff --git a/kernel/dma/Kconfig b/kernel/dma/Kconfig
index a0ce3c1494fdd2..cb05247804d9f3 100644
--- a/kernel/dma/Kconfig
+++ b/kernel/dma/Kconfig
@@ -1,5 +1,8 @@
 # SPDX-License-Identifier: GPL-2.0-only
 
+config NO_DMA
+	bool
+
 config HAS_DMA
 	bool
 	depends on !NO_DMA
-- 
2.26.2

[PATCH 09/10] sh: don't allow non-coherent DMA for NOMMU

[Christoph Hellwig]

The code handling non-coherent DMA depends on being able to remap code
as non-cached.  But that can't be done without an MMU, so using this
option on NOMMU builds is broken.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/sh/Kconfig | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/sh/Kconfig b/arch/sh/Kconfig
index f8027eee08edae..337eb496c45a0a 100644
--- a/arch/sh/Kconfig
+++ b/arch/sh/Kconfig
@@ -61,6 +61,7 @@ config SUPERH
 	select MAY_HAVE_SPARSE_IRQ
 	select MODULES_USE_ELF_RELA
 	select NEED_SG_DMA_LENGTH
+	select NO_DMA if !MMU && !DMA_COHERENT
 	select NO_GENERIC_PCI_IOPORT_MAP if PCI
 	select OLD_SIGACTION
 	select OLD_SIGSUSPEND
@@ -135,7 +136,7 @@ config DMA_COHERENT
 	bool
 
 config DMA_NONCOHERENT
-	def_bool !DMA_COHERENT
+	def_bool !NO_DMA && !DMA_COHERENT
 	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
 
 config PGTABLE_LEVELS
-- 
2.26.2

[PATCH 10/10] sh: use the generic dma coherent remap allocator

[Christoph Hellwig]

This switches to using common code for the DMA allocations, including
potential use of the CMA allocator if configured.

Switching to the generic code enables DMA allocations from atomic
context, which is required by the DMA API documentation, and also
adds various other minor features drivers start relying upon.  It
also makes sure we have on tested code base for all architectures
that require uncached pte bits for coherent DMA allocations.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/sh/Kconfig               |  2 ++
 arch/sh/kernel/dma-coherent.c | 51 ++---------------------------------
 2 files changed, 4 insertions(+), 49 deletions(-)

diff --git a/arch/sh/Kconfig b/arch/sh/Kconfig
index 337eb496c45a0a..32d959849df9bc 100644
--- a/arch/sh/Kconfig
+++ b/arch/sh/Kconfig
@@ -137,7 +137,9 @@ config DMA_COHERENT
 
 config DMA_NONCOHERENT
 	def_bool !NO_DMA && !DMA_COHERENT
+	select ARCH_HAS_DMA_PREP_COHERENT
 	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
+	select DMA_DIRECT_REMAP
 
 config PGTABLE_LEVELS
 	default 3 if X2TLB
diff --git a/arch/sh/kernel/dma-coherent.c b/arch/sh/kernel/dma-coherent.c
index d4811691b93cc1..cd46a9825e3c59 100644
--- a/arch/sh/kernel/dma-coherent.c
+++ b/arch/sh/kernel/dma-coherent.c
@@ -3,60 +3,13 @@
  * Copyright (C) 2004 - 2007  Paul Mundt
  */
 #include <linux/mm.h>
-#include <linux/init.h>
 #include <linux/dma-noncoherent.h>
-#include <linux/module.h>
 #include <asm/cacheflush.h>
 #include <asm/addrspace.h>
 
-void *arch_dma_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle,
-		gfp_t gfp, unsigned long attrs)
+void arch_dma_prep_coherent(struct page *page, size_t size)
 {
-	void *ret, *ret_nocache;
-	int order = get_order(size);
-
-	gfp |= __GFP_ZERO;
-
-	ret = (void *)__get_free_pages(gfp, order);
-	if (!ret)
-		return NULL;
-
-	/*
-	 * Pages from the page allocator may have data present in
-	 * cache. So flush the cache before using uncached memory.
-	 */
-	arch_sync_dma_for_device(virt_to_phys(ret), size,
-			DMA_BIDIRECTIONAL);
-
-	ret_nocache = (void __force *)ioremap(virt_to_phys(ret), size);
-	if (!ret_nocache) {
-		free_pages((unsigned long)ret, order);
-		return NULL;
-	}
-
-	split_page(pfn_to_page(virt_to_phys(ret) >> PAGE_SHIFT), order);
-
-	*dma_handle = virt_to_phys(ret);
-	if (!WARN_ON(!dev))
-		*dma_handle -= PFN_PHYS(dev->dma_pfn_offset);
-
-	return ret_nocache;
-}
-
-void arch_dma_free(struct device *dev, size_t size, void *vaddr,
-		dma_addr_t dma_handle, unsigned long attrs)
-{
-	int order = get_order(size);
-	unsigned long pfn = (dma_handle >> PAGE_SHIFT);
-	int k;
-
-	if (!WARN_ON(!dev))
-		pfn += dev->dma_pfn_offset;
-
-	for (k = 0; k < (1 << order); k++)
-		__free_pages(pfn_to_page(pfn + k), 0);
-
-	iounmap(vaddr);
+	__flush_purge_region(page_address(page), size);
 }
 
 void arch_sync_dma_for_device(phys_addr_t paddr, size_t size,
-- 
2.26.2

Re: [PATCH 09/10] sh: don't allow non-coherent DMA for NOMMU

[Rob Landley]

On 6/26/20 3:07 AM, Christoph Hellwig wrote:
> The code handling non-coherent DMA depends on being able to remap code
> as non-cached.  But that can't be done without an MMU, so using this
> option on NOMMU builds is broken.

I'm working on a nommu j-core board that's doing DMA behind the OS's back at the
moment, which I have a todo item to teach the kernel about. The DMA does not go
through the cache, there's currently a cache flush before looking at the result
instead.

How should this be wired up after your patch?

Rob

Re: [PATCH 09/10] sh: don't allow non-coherent DMA for NOMMU

[Christoph Hellwig]

On Sat, Jun 27, 2020 at 08:01:17PM -0500, Rob Landley wrote:
> On 6/26/20 3:07 AM, Christoph Hellwig wrote:
> > The code handling non-coherent DMA depends on being able to remap code
> > as non-cached.  But that can't be done without an MMU, so using this
> > option on NOMMU builds is broken.
> 
> I'm working on a nommu j-core board that's doing DMA behind the OS's back at the
> moment, which I have a todo item to teach the kernel about. The DMA does not go
> through the cache, there's currently a cache flush before looking at the result
> instead.
> 
> How should this be wired up after your patch?

The problem with nommu and non-coherent dma is the dma_alloc_coherent
calls.  Most platforms with an mmu set a nocache bit through the page
tables (including sh), but that option obviously doesn't exist for
nommu.  Some hardware has an uncached window where access is uncached
automatically if access through specific kernel virtual addresses,
for that the architecture needs to impement the arch_dma_set_uncached
helper and select CONFIG_ARCH_HAS_DMA_SET_UNCACHED.  If that also
doesn't exist you'll need some sort of pool of always uncached
memory (set by the firmware or early startup code).  That currently
doesn't exist in generic code, but we have a bunch of architectures
implementing that in arch_dma_alloc.  I plan to have a common
implementation of the pool soon hopefully.

Streaming DMA just works if you reuse the existing
arch_sync_dma_for_device implementation.

Re: [PATCH 09/10] sh: don't allow non-coherent DMA for NOMMU

[Ulf Hansson]

On Sat, 29 Aug 2020 at 10:31, Christoph Hellwig <hch@lst.de> wrote:
>
> On Fri, Aug 28, 2020 at 11:09:43AM -0400, Rich Felker wrote:
> > > However, by looking at the code, one get the feeling that the DMA
> > > support is somewhat prepared to be made optional. I guess it has never
> > > been really tested, as the Kconfig option has "depends on HAS_DMA"  -
> > > and it's been like that as long as I can remember.
> >
> > It always worked on our "byte-banged" SPI controller, with no DMA
> > controller present, before Christoph's changes in this patch series,
>
> Before that nommu sh builds provided a DMA mapping implementation
> that even worked for the streaming side (dma_map_*), but would corrupt
> data if you used dma_alloc_coherent memory to communicate with the
> device.

I see.

>
> > and seems to be working now (although I have some other, hopefully
> > unrelated regressions to debug) with #ifdef CONFIG_HAS_DMA around the
> > if (spi->master->dev.parent->dma_mask) block in mmc_spi_probe. That's
> > probably not the right fix though -- why isn't it checking
> > host->dma_dev instead and only attempting DMA setup if dma_dev is
> > non-null?
>
> I don't think dma_dev can be NULL right now.  dma_dev is assigned here:
>
>         if (spi->master->dev.parent->dma_mask) {
>                 struct device   *dev = spi->master->dev.parent;
>
>                 host->dma_dev = dev;
>
> but for any OF or real bus device dma_mask never is zero (it actually is
> a pointer), and the value of it also is initialized to 32-bit by default,
> making this effectively an "if (1) {".  The driver needs some way to
> communicate if a given device actually is DMA capable or not. Or is that
> purely a factor of the platform which would be a little strange.

Okay, thanks for clarifying things.

>
> In which case we should do something like:
>
>
> diff --git a/drivers/mmc/host/mmc_spi.c b/drivers/mmc/host/mmc_spi.c
> index 39bb1e30c2d722..3b0cc9a70e6432 100644
> --- a/drivers/mmc/host/mmc_spi.c
> +++ b/drivers/mmc/host/mmc_spi.c
> @@ -1374,7 +1374,7 @@ static int mmc_spi_probe(struct spi_device *spi)
>         if (!host->data)
>                 goto fail_nobuf1;
>
> -       if (spi->master->dev.parent->dma_mask) {
> +       if (IS_ENABLED(CONFIG_HAS_DMA)) {

I guess "#ifdef CONFIG_HAS_DMA", would work as well as CONFIG_HAS_DMA
can't be "m".

I would also suggest moving the DMA setup things into a separate
function. Then when CONFIG_HAS_DMA is unset, we can use a stub.

>                 struct device   *dev = spi->master->dev.parent;
>
>                 host->dma_dev = dev;

Let me cook a patch that we can try.

Kind regards
Uffe

Re: [PATCH 09/10] sh: don't allow non-coherent DMA for NOMMU

[Christoph Hellwig]

On Fri, Aug 28, 2020 at 11:09:43AM -0400, Rich Felker wrote:
> > However, by looking at the code, one get the feeling that the DMA
> > support is somewhat prepared to be made optional. I guess it has never
> > been really tested, as the Kconfig option has "depends on HAS_DMA"  -
> > and it's been like that as long as I can remember.
> 
> It always worked on our "byte-banged" SPI controller, with no DMA
> controller present, before Christoph's changes in this patch series,

Before that nommu sh builds provided a DMA mapping implementation
that even worked for the streaming side (dma_map_*), but would corrupt
data if you used dma_alloc_coherent memory to communicate with the
device. 

> and seems to be working now (although I have some other, hopefully
> unrelated regressions to debug) with #ifdef CONFIG_HAS_DMA around the
> if (spi->master->dev.parent->dma_mask) block in mmc_spi_probe. That's
> probably not the right fix though -- why isn't it checking
> host->dma_dev instead and only attempting DMA setup if dma_dev is
> non-null?

I don't think dma_dev can be NULL right now.  dma_dev is assigned here:

	if (spi->master->dev.parent->dma_mask) {
		struct device   *dev = spi->master->dev.parent;

		host->dma_dev = dev;

but for any OF or real bus device dma_mask never is zero (it actually is
a pointer), and the value of it also is initialized to 32-bit by default,
making this effectively an "if (1) {".  The driver needs some way to
communicate if a given device actually is DMA capable or not. Or is that
purely a factor of the platform which would be a little strange.

In which case we should do something like:


diff --git a/drivers/mmc/host/mmc_spi.c b/drivers/mmc/host/mmc_spi.c
index 39bb1e30c2d722..3b0cc9a70e6432 100644
--- a/drivers/mmc/host/mmc_spi.c
+++ b/drivers/mmc/host/mmc_spi.c
@@ -1374,7 +1374,7 @@ static int mmc_spi_probe(struct spi_device *spi)
 	if (!host->data)
 		goto fail_nobuf1;
 
-	if (spi->master->dev.parent->dma_mask) {
+	if (IS_ENABLED(CONFIG_HAS_DMA)) {
 		struct device	*dev = spi->master->dev.parent;
 
 		host->dma_dev = dev;

Re: [PATCH 09/10] sh: don't allow non-coherent DMA for NOMMU

[Rich Felker]

On Fri, Aug 28, 2020 at 11:26:57AM +0200, Ulf Hansson wrote:
> On Fri, 28 Aug 2020 at 06:24, Christoph Hellwig <hch@lst.de> wrote:
> >
> > On Thu, Aug 27, 2020 at 10:11:53PM -0400, Rich Felker wrote:
> > > > This change broke SD card support on J2 because MMC_SPI spuriously
> > > > depends on HAS_DMA. It looks like it can be fixed just by removing
> > > > that dependency from drivers/mmc/host/Kconfig.
> > >
> > > It can't. mmp_spi_probe fails with ENOMEM, probably due to trying to
> > > do some DMA setup thing that's not going to be needed if the
> > > underlying SPI device doesn't support/use DMA.
> >
> > Adding the linux-mmc and linux-spi lists, as that seems pretty odd.
> 
> The mmc_spi driver needs modernizations, so I am not surprised to see
> odd things.
> 
> My guess is that in ->probe() we check "if
> (spi->master->dev.parent->dma_mask)" - > and runs dma_map*()
> operations, which fails and leads to bailing out of ->probe() to
> return an error code.
> 
> However, by looking at the code, one get the feeling that the DMA
> support is somewhat prepared to be made optional. I guess it has never
> been really tested, as the Kconfig option has "depends on HAS_DMA"  -
> and it's been like that as long as I can remember.

It always worked on our "byte-banged" SPI controller, with no DMA
controller present, before Christoph's changes in this patch series,
and seems to be working now (although I have some other, hopefully
unrelated regressions to debug) with #ifdef CONFIG_HAS_DMA around the
if (spi->master->dev.parent->dma_mask) block in mmc_spi_probe. That's
probably not the right fix though -- why isn't it checking
host->dma_dev instead and only attempting DMA setup if dma_dev is
non-null?

Rich

Re: [PATCH 09/10] sh: don't allow non-coherent DMA for NOMMU

[Ulf Hansson]

On Fri, 28 Aug 2020 at 06:24, Christoph Hellwig <hch@lst.de> wrote:
>
> On Thu, Aug 27, 2020 at 10:11:53PM -0400, Rich Felker wrote:
> > > This change broke SD card support on J2 because MMC_SPI spuriously
> > > depends on HAS_DMA. It looks like it can be fixed just by removing
> > > that dependency from drivers/mmc/host/Kconfig.
> >
> > It can't. mmp_spi_probe fails with ENOMEM, probably due to trying to
> > do some DMA setup thing that's not going to be needed if the
> > underlying SPI device doesn't support/use DMA.
>
> Adding the linux-mmc and linux-spi lists, as that seems pretty odd.

The mmc_spi driver needs modernizations, so I am not surprised to see
odd things.

My guess is that in ->probe() we check "if
(spi->master->dev.parent->dma_mask)" - > and runs dma_map*()
operations, which fails and leads to bailing out of ->probe() to
return an error code.

However, by looking at the code, one get the feeling that the DMA
support is somewhat prepared to be made optional. I guess it has never
been really tested, as the Kconfig option has "depends on HAS_DMA"  -
and it's been like that as long as I can remember.

Kind regards
Uffe

Re: [PATCH 09/10] sh: don't allow non-coherent DMA for NOMMU

[Christoph Hellwig]

On Thu, Aug 27, 2020 at 10:11:53PM -0400, Rich Felker wrote:
> > This change broke SD card support on J2 because MMC_SPI spuriously
> > depends on HAS_DMA. It looks like it can be fixed just by removing
> > that dependency from drivers/mmc/host/Kconfig.
> 
> It can't. mmp_spi_probe fails with ENOMEM, probably due to trying to
> do some DMA setup thing that's not going to be needed if the
> underlying SPI device doesn't support/use DMA.

Adding the linux-mmc and linux-spi lists, as that seems pretty odd.

Re: [PATCH 09/10] sh: don't allow non-coherent DMA for NOMMU

[Rich Felker]

On Thu, Aug 27, 2020 at 10:00:48PM -0400, Rich Felker wrote:
> On Tue, Jul 14, 2020 at 02:18:55PM +0200, Christoph Hellwig wrote:
> > The code handling non-coherent DMA depends on being able to remap code
> > as non-cached.  But that can't be done without an MMU, so using this
> > option on NOMMU builds is broken.
> > 
> > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > ---
> >  arch/sh/Kconfig | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> > 
> > diff --git a/arch/sh/Kconfig b/arch/sh/Kconfig
> > index f8027eee08edae..337eb496c45a0a 100644
> > --- a/arch/sh/Kconfig
> > +++ b/arch/sh/Kconfig
> > @@ -61,6 +61,7 @@ config SUPERH
> >  	select MAY_HAVE_SPARSE_IRQ
> >  	select MODULES_USE_ELF_RELA
> >  	select NEED_SG_DMA_LENGTH
> > +	select NO_DMA if !MMU && !DMA_COHERENT
> >  	select NO_GENERIC_PCI_IOPORT_MAP if PCI
> >  	select OLD_SIGACTION
> >  	select OLD_SIGSUSPEND
> > @@ -135,7 +136,7 @@ config DMA_COHERENT
> >  	bool
> 
> This change broke SD card support on J2 because MMC_SPI spuriously
> depends on HAS_DMA. It looks like it can be fixed just by removing
> that dependency from drivers/mmc/host/Kconfig.

It can't. mmp_spi_probe fails with ENOMEM, probably due to trying to
do some DMA setup thing that's not going to be needed if the
underlying SPI device doesn't support/use DMA.

Rich

Re: [PATCH 09/10] sh: don't allow non-coherent DMA for NOMMU

[Rich Felker]

On Tue, Jul 14, 2020 at 02:18:55PM +0200, Christoph Hellwig wrote:
> The code handling non-coherent DMA depends on being able to remap code
> as non-cached.  But that can't be done without an MMU, so using this
> option on NOMMU builds is broken.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  arch/sh/Kconfig | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/sh/Kconfig b/arch/sh/Kconfig
> index f8027eee08edae..337eb496c45a0a 100644
> --- a/arch/sh/Kconfig
> +++ b/arch/sh/Kconfig
> @@ -61,6 +61,7 @@ config SUPERH
>  	select MAY_HAVE_SPARSE_IRQ
>  	select MODULES_USE_ELF_RELA
>  	select NEED_SG_DMA_LENGTH
> +	select NO_DMA if !MMU && !DMA_COHERENT
>  	select NO_GENERIC_PCI_IOPORT_MAP if PCI
>  	select OLD_SIGACTION
>  	select OLD_SIGSUSPEND
> @@ -135,7 +136,7 @@ config DMA_COHERENT
>  	bool

This change broke SD card support on J2 because MMC_SPI spuriously
depends on HAS_DMA. It looks like it can be fixed just by removing
that dependency from drivers/mmc/host/Kconfig.

Rich

[PATCH 09/10] sh: don't allow non-coherent DMA for NOMMU

[Christoph Hellwig]

The code handling non-coherent DMA depends on being able to remap code
as non-cached.  But that can't be done without an MMU, so using this
option on NOMMU builds is broken.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/sh/Kconfig | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/sh/Kconfig b/arch/sh/Kconfig
index f8027eee08edae..337eb496c45a0a 100644
--- a/arch/sh/Kconfig
+++ b/arch/sh/Kconfig
@@ -61,6 +61,7 @@ config SUPERH
 	select MAY_HAVE_SPARSE_IRQ
 	select MODULES_USE_ELF_RELA
 	select NEED_SG_DMA_LENGTH
+	select NO_DMA if !MMU && !DMA_COHERENT
 	select NO_GENERIC_PCI_IOPORT_MAP if PCI
 	select OLD_SIGACTION
 	select OLD_SIGSUSPEND
@@ -135,7 +136,7 @@ config DMA_COHERENT
 	bool
 
 config DMA_NONCOHERENT
-	def_bool !DMA_COHERENT
+	def_bool !NO_DMA && !DMA_COHERENT
 	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
 
 config PGTABLE_LEVELS
-- 
2.26.2

[PATCH 09/10] sh: don't allow non-coherent DMA for NOMMU

[Christoph Hellwig]

The code handling non-coherent DMA depends on being able to remap code
as non-cached.  But that can't be done without an MMU, so using this
option on NOMMU builds is broken.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/sh/Kconfig | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/sh/Kconfig b/arch/sh/Kconfig
index 808b638b11f3..b97d9712c252 100644
--- a/arch/sh/Kconfig
+++ b/arch/sh/Kconfig
@@ -48,6 +48,7 @@ config SUPERH
 	select MAY_HAVE_SPARSE_IRQ
 	select MODULES_USE_ELF_RELA
 	select NEED_SG_DMA_LENGTH
+	select NO_DMA if !MMU && !DMA_COHERENT
 	select NO_GENERIC_PCI_IOPORT_MAP if PCI
 	select OLD_SIGACTION
 	select OLD_SIGSUSPEND
@@ -155,7 +156,7 @@ config DMA_COHERENT
 	bool
 
 config DMA_NONCOHERENT
-	def_bool !DMA_COHERENT
+	def_bool !NO_DMA && !DMA_COHERENT
 	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
 
 config PGTABLE_LEVELS
-- 
2.25.1